Multi-omics analysis reveals an anoikis-related signature for non-small cell lung cancer

Multi-omics analysis reveals an anoikis-related signature for non-small cell lung cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Multi-omics analysis reveals an anoikis-related signature for non-small cell lung cancer Yuqi Ma, Jia Li, Tao Shen This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4640324/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Non-small cell lung cancer (NSCLC) is a prevalent form of lung cancer characterized by a significant death rate. Anoikis (ANO), refers to a distinct kind of programmed cell death that is strongly linked to the body's immune response to cancer. Nevertheless, the precise function of ANO in NSCLC is still not well understood. Methods ANO-related genes were analysed using multiple methods, including AUCell, UCell, single-sample gene set enrichment analysis (ssGSEA), Singscore, AddModuleScore, GSVA and weighted gene co-expression network analysis (WGCNA). We have developed an innovative machine learning framework that combines 10 different machine learning algorithms and 101 possible combinations of these algorithms. The goal of this framework is to build a reliable signature, known as the Anoikis-related signature (ARS), which is related to the phenomenon of anoikis. The performance of ARS was evaluated in both the training and validation sets. Column line graphs using ARS were developed as a quantitative technique to predict prognosis in clinical settings. Multi-omics studies, including genomic and bulk transcriptomic, were performed to gain more in-depth knowledge of prognostic features. We analysed the responsiveness of risk groups to immunotherapy and searched for tailored drugs to target specific risk categories. Results We discovered 103 genes associated with ANO at both single cell and bulk transcriptome levels. A computational framework using machine learning and 101 combinations was used to generate the consensus ARS. This framework showed exceptional performance in accurately predicting prognosis and clinical change, and the ARS can also be used to predict the initiation, progression and spread of NSCLC. Statistical studies have shown that it is an independent prognostic determinant of (OS) and disease-specific survival (DSS) in NSCLC. The integrated column line graphs of the ARS provide an accurate and quantitative tool for clinical practice. We also identified distinct metabolic processes, patterns of genetic mutations and the presence of immune cells in the tumour microenvironment that differed between the high-risk and low-risk groups. Significantly, there were significant changes in the immunophenotype score (IPS) between the risk groups, suggesting that the high-risk group is likely to have a more favourable response to immunotherapy. In addition, potential drugs targeting specific at-risk populations were identified. Conclusion The purpose of our work is to create a signature associated with immunogenic cell death. This signature has the potential to be a useful tool for predicting the prognosis of NSCLC, as well as for targeted prevention and personalised therapy. We are also providing new insights into the molecular pathways involved in the growth and progression of NSCLC through the use of mass transcriptomics and genomics research. Biological sciences/Cancer Biological sciences/Computational biology and bioinformatics NSCLC single-cell machine learning anoikis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Introduction Accounting for 18% of all cancer deaths worldwide, lung cancer is the leading cause of cancer deaths( 1 ). Non-small cell lung cancer (NSCLC), with an estimated 5-year survival rate of 10–23%, is the most common type of lung cancer( 2 ). NSCLC is typically diagnosed at a late stage, which is a major limitation to the effectiveness of treatment( 3 ). As a result, there is an urgent need for innovative biomarkers to help predict the prognosis of patients with NSCLC. These biomarkers would allow timely clinical intervention to slow disease progression. Immune checkpoint inhibitors (ICIs) have been generally accepted by patients and have changed the management of advanced NSCLC. ICIs stimulate the body's immune system to detect and remove cancer cells( 4 ). However, these immune-based therapies only assist a fraction of NSCLC patients because primary resistance remains prominent in many situations ( 5 , 6 ). In recent years, extensive studies of various cell death modalities have identified cell death as a critical mechanism that can be exploited in therapy. Anoikis is a form of programmed cell death that induces cell death primarily by disrupting endogenous mitochondrial pathways or activating exogenous pathways through cell surface death receptors( 7 , 8 ). Therefore, the identification of biomarkers or characteristics that are associated with prognosis and response to ICIs in patients with NSCLC is essential. Recently, a growing number of anoikis-related signature (ARS) have been identified in various cancers. Ye et al. showed that MYH9 expression is upregulated in metastatic gastric cancer tissue and is associated with poor prognosis in gastric cancer patients. Further studies showed that MYH9 binds to the CTNNB1 promoter through its DNA binding domain, conferring resistance to anoikis in tumour cells( 9 ). GDH1-mediated reprogramming of glutamine catabolism has been shown to be a driver of anoikis resistance and tumour metastasis in LKB1-deficient lung cancer( 10 ). In addition, ARS have been shown to be associated with prognosis in several tumours( 11 , 12 ). However, comprehensive analyses of the impact of anoikis in NSCLC remain scarce. For instance, the use of biomarkers associated with anoikis to stratify patients according to prognosis and to predict their response to immunotherapy and chemotherapy. This is an area of significant importance for future studies. This research intended to examine the features of ARS at various histological levels. We turned both single-cell and bulk transcriptomes to find ARS. To generate consistent ARS, we deployed a unique machine learning framework that consisted of 10 machine learning methods and 101 combinations. In order to improve the implementation of ARS, we conducted an assessment to see whether ARS could accurately forecast the incidence, progression, and spread of NSCLC. In addition, we developed a nomogram that incorporates ARS to offer a precise and statistical approach for estimating patient prognosis in clinical settings. We also analysed how different at-risk subgroups responded to treatment with ICIs and their sensitivity to first-line NSCLC therapies such as tyrosine kinase inhibitors (TKIs) and cytotoxic drugs. Our goal is to personalize prognostic prediction and treatment for NSCLC patients. See the flow chart (Fig. 1) for details. Methods Data collection and processing We obtained RNA-seq data and accompanying clinical data for NSCLC from The Cancer Genome Atlas (TCGA) and isolated transcripts per million (TPM) values for further analysis. Samples with incomplete or missing clinicopathological information were excluded, resulting in a dataset of 501 TCGA patients. In addition, we utilized the NSCLC single-cell RNA sequencing dataset GSE131907 from the Gene Expression Omnibus (GEO) collection. The Mutation Annotation Format (MAF) was used to extract somatic mutation data from TCGA. Copy number variation (CNV) data for TCGA-LUAD patients were collected from the UCSC Xena database. To assess the value of prognostic features for forecasting patient outcome following immunotherapy, we included the IMvigor210 cohort. ( 13 ). In addition, to examine the prediction ability of these indicators for NSCLC incidence, we added two datasets from the GEO database (GSE29013 and GSE42127). To detect ARS, we gathered 338 these genes from previous research.( 12 , 14 , 15 ). (Supplementary file 1) scRNA analysis We obtained scRNA sequencing data from 44 patients (58 samples) from the GSE131907. We analysed the single-cell sequencing data using the Seurat package. Quality control (QC) process included selecting cells with mitochondrial gene content below 20% and genes that were expressed in no less than three cells within a range of 500 to 7000 in terms of expression levels. Highly variable genes were identified for subsequent analysis with a variable gene number of 2000. The Harmony package suite was used to remove batch effects in the data from the 58 samples. The FindClusters and FindNeighbors routines were used to generate cell clusters, which were then visualized using the tSNE approach. Ultimately, cellular annotation was conducted using marker genes specific to various cell types. The expression levels of anoikis (ANO) in all cells were determined using the five techniques of AUCell, UCell, GSVA, Singscore and Addmodulescore. The total score for all cell types was then determined to quantify the activity of ANO in all cell types. Among the eight cell types, we found that epithelial cells had significantly increased ANO activity. We divided the cells into high and low ANO groups. We then looked for genes that were differentially expressed in the two groups so that we could study them further. ssGSEA Single-sample gene set enrichment analysis (ssGSEA) has become a commonly used technique for measuring the enrichment scores of certain gene sets in individual samples. The ssGSEA score of each sample reflects the degree of systematic up- or down-regulation of a specific gene set in the sample. We calculated the ANO score for each TCGA-LUAD sample by using ssGSEA in the R package "GSVA" in our study( 16 ). In order to determine the possible pathways related to the attributes, we conducted gene set enrichment analysis (GSVA) to compute scores for 50 signature pathways. We next analyzed the pathways that showed significant differences between the high-risk and low-risk groups using the 'limma' package( 17 ). Furthermore, to clarify the biological processes (BP), cellular components (CC), and molecular functions (MF) associated with the various risk subgroups, we used the R package 'clusterProfiler' with a false discovery rate (FDR) threshold of less than 0.25 and an absolute normalized enrichment score (NES) greater than 1. WGCNA WGCNA is a methodical biological technique that examines patterns of genetic linkages among various samples and discovers genomes that are highly synergistically changed. We used the WGCNA package to conduct a WGCNA analysis using the TCGA-LUAD dataset in our research. First, an acceptable soft threshold β was established to meet the conditions for designing a scale-free network. A dynamic tree-cutting method was applied for gene clustering and module identification. Finally, the module having the strongest connection with ANO scores were picked for further analysis. Constructing prognostic models by integrating several machine learning approaches We utilized the limma package to undertake a difference analysis between normal and tumour samples in the TCGA-LUAD dataset. Subsequently, we performed bulk RNA-seq sequencing of differentially expressed genes (DEGs) that were identified by WGCNA inside modules associated with anoikis. We designated these sequenced genes as ARS because they were shown to be associated with anoikis in both the bulk and single-cell transcriptome analyses. In order to create resilient prognostic models with exceptional predicted precision, we adhered to the following procedures: a: Initially, we used univariate Cox regression analysis to identify ARS that may have predictive significance in the TCGA-LUAD dataset. b: The TCGA-LUAD dataset was split into two sets at random, one for internal validation and the other for training, in order to guarantee a fair distribution of clinical characteristics across the two sets. There are ten machine learning algorithms: GBM, Lasso, Ridge, Stepwise Cox, CoxBoost, Random Survival Forest (RSF), Elastic Net (Enet), Supervised Principal Component (SuperPC), Partial Least Squares Regression Cox (plsRcox), and Survival Support Vector Machine (Survival-SVM). Using a tenfold cross-validation framework, we organized 101 combinations of these 10 techniques for variable selection and model creation in the TCGA-LUAD training dataset. c: Two groups were randomly created from the TCGA-LUAD dataset: one for internal validation and one for training. The following 10 machine learning techniques were available: Enet, SuperPC, plsRcox, RSF, Lasso, Ridge, Stepwise Cox, CoxBoost, GBM, and SVM. We selected variables and built models for 101 combinations of these 10 approaches in the TCGA-LUAD training dataset using a tenfold cross-validation framework. Predictive nomogram formation and survival analysis Based on the median ARS risk score, the TCGA training set was split into high-risk and low-risk groups. Then, using the R package "survminer," we did KM curve analysis to see whether the high- and low-risk groups' OS and DSS differed significantly from one another (p < 0.05). In addition, we evaluated the sensitivity and specificity of ARS in predicting OS in NSCLC patients by doing ROC curve analysis using the "timeROC" software program. We also contrasted other clinical features with AUC of ARS. Furthermore, we investigated the relationship between ARS and a number of clinical attributes, such as grade, age, gender, stage, T, M, and N. Using TCGA-LUAD, we ran univariate and multivariate Cox regression analysis to see whether ARS was a reliable predictor of survival for patients with non-small cell lung cancer. Our goal was to increase the prediction capacity and prognostic accuracy of our model. To do this, we created a nomogram that quantified the predicted survival of NSCLC patients by combining clinical characteristics and ARS. Finally, we assessed the precision and accuracy of the nomogram using ROC curves, C-indices, and calibration curves, as well as its net clinical benefit using decision curve analysis (DCA). Analysis of genomic variation between ARS risk subgroups Mutant Allele Tumor Heterogeneity (MATH) is a method of quantifying ITH based on the distribution of mutant alleles. MATH scores provide a quantifiable and measurable assessment of ITH( 18 ). Research has investigated the predictive relevance of MATH in a range of tumour types( 19 ), including breast, colorectal and head and neck malignancies( 20 , 21 ). For this investigation, we calculated each NSCLC patient's MATH score using a previously published method. Subsequently, we conducted survival analyses by using their MATH score. In order to look at somatic mutations linked to ARS, we created waterfall plots using the maftools package, which show the mutational landscape of NSCLC patients in high- and low-risk groups. We also performed copy number variation (CNV) analysis on 30 genes that showed the most variation between the high and low-risk groups. An extensive perspective of immunological characteristics and responsiveness to immunological checkpoint inhibitor therapy One important component of tumor immunotherapy is the anti-cancer immune cycle, which consists of seven stages: releasing the cancer antigen (step1), presenting the cancer antigen (step2), triggering and activation (step3), transporting immune cells to the tumour (step4), infiltrating the tumour (step5), having T-cells recognize the tumour (step6), and killing the tumour (step7). We quantified the infiltration of 22 immune cells using the CIBERSORT method in order to examine the connection between ARS and immune cell infiltration in the NSCLC tumour microenvironment. We also used the ESTIMATE and ssGSEA to verify the accuracy of the CIBERSORT findings. We used the Immunophenotype Scoring (IPS) system to predict immunogenicity based on immunomodulators, immunosuppressive cells, MHC molecules, and effector cells. The IPS system uses a machine learning approach to build IPS scores based on unbiased gene expression of representative cell types. Higher IPS scores are indicative with a better immunotherapy response. The IPS ratings for patient samples from TCGA-LUAD were obtained using the Cancer Immunome Atlas (TCIA) database. ( https://tcia.at/at-home ) Importance of the ARS for medication sensitivity To individualise treatment, we predicted chemotherapy sensitivity in colorectal cancer patients with different ARS risk scores using the R package pRRophetic( 22 ). By employing pRRophetic to match patient tissue gene expression patterns to cancer cell line patterns, half-maximal inhibitory concentrations (IC50) were found. The difference in medication IC50 between the high and low risk groups was investigated using the Wilcoxon test. When p < 0.05, it was deemed statistically significant. We also used the Connection Map of Gene Expression Profiles (CMap) database to evaluate substances that might be activating or inhibiting agents. According to Yang et al( 23 ). web-based studies of CMap databases are less likely to yield relevant results than CMap analyses using the R-based "extreme sum" (XSum) technique. We used the XSum method to determine CMap scores. If the drug has a CMap score between − 1 and 0, it may be able to help people with ccRCC by acting as a cancer drug. Statistical analysis R software (versions: R 4.3.2) was used for all statistical analyses. The chi-squared test was used to analyse differences in clinical characteristics between the internal validation and training sets. The Wilcoxon test was a non-parametric method for estimating the difference between two non-normally distributed variables. Significant changes in DEGs were evaluated using FDR-corrected p-values. To investigate independent prognostic factors, R was used to conduct both univariate and multivariate Cox regression analysis. ROC curve analysis and AUC were computed using the "timeROC" package to evaluate the model's efficacy. The association between immune cell infiltration and risk score was examined using Spearman's correlation analysis. Results Characteristics of anoikis in single cell transcriptome We acquired 208,506 cells from 44 NSCLC patients' single-cell RNA sequencing data. We successfully integrated 58 samples using the Harmony software program in order to remove batch effects, as shown in Fig. 2A . The top 2000 variant genes were then subjected to downscaling using principal component analysis (PCA) and t-distributed stochastic neighbour embedding (t-SNE). After that, at a resolution of 1.5, clustering was done to combine all of the cells into 54 clusters (Supplementary file 2) . We classified the cells into eight main clusters, NK cell, B lymphocytes, T lymphocytes, Epithelial cell, Fibroblasts, Endothelial cell, Myeloid cell, Mast cell using flag genes specific to various celltypes. The top 5 marker genes for each cell group are shown on the heatmap (Fig. 2B) . The expression levels of anoikis (ANO) in all cells were determined using the five techniques of AUCell, UCell, GSVA, Singscore and Addmodulescore (Fig. 2D) . The total score for all cell types was then determined to quantify the activity of ANO in all cell types (Fig. 2C) . Of the eight cell types, we found a significant increase in ANO activity in epithelial cells, followed by fibroblast cells. Based on ANO activity, we divided the cells into high and low ANO groups. We then looked for genes that were differentially expressed in the two groups so that we could study them further. Bulk RNA sequencing to identify genes and the hub module associated with ANO. ssGSEA When evaluating alterations in biological processes and pathway activity in particular samples, the ssGSEA approach is often used. For each TCGA-LUAD sample in this investigation, we calculated ANO activity scores using the ssGSEA method. These scores served as phenotypic data for further WGCNA analysis. We used WGCNA analysis on the TCGA-LUAD dataset to find modules that were significantly associated with ANO scores. More specifically, co-expression networks were constructed by identifying ARS at the single-cell sequencing level after eliminating outlier data (Fig. 3A) . To ensure scale-free topological networks, power = 5 was chosen as the ideal soft threshold (Supplementary file 3) . By adjusting MEDissThres to 0.25 and the minimum module gene number to 60, a total of four modules were found (Fig. 3B) . The brown module and the ANO score in batch RNA-seq showed a good correlation according to our data (cor = 0.61, Fig. 3C). The TCGA-LUAD bulk RNA-seq volcano plot showed DEGs between tumour and normal tissues (p < 0.05) (Fig. 3D) . After using DEG from bulk RNA-seq to sequence 229 genes in the brown module, we were able to identify a total of 103 genes (Fig. 3E) . These genes were termed ARS, and it is thought that both the whole transcriptome and the transcriptome of a particular cell are affected by ANO. Biological processes (BPs) were found to be significantly enriched in the GO enrichment analysis of ARS (Fig. 3F) . In addition, we conducted an analysis on the frequency of CNV for the 36 genes. Among them, GMFG exhibited a copy number increase with a frequency above 10%, as seen in Fig. 3G . The protein-protein interaction (PPI) network of these genes is shown in Supplementary file 4 . Prognostic feature construction with integrated machine learning Using a mixture of 101 machine learning approaches, we analyzed 31 prognostic genes obtained from univariate Cox regression analysis in order to provide consistent ARS. As shown in Fig. 4H , using a tenfold cross-validation framework, we fitted 101 prediction models to the training set and determined C-indexes for both the training and validation sets. The prediction models with the greatest average C-index among all the models were constructed using the RSF and Lasso algorithms after the analysis of 101 models. Within the training, internal, and external validation sets, this model demonstrated strong prediction performance. After careful screening, we discovered that Lasso + RSF is a prediction model with outstanding accuracy and translational relevance. By minimising the biased likelihood and using a tenfold cross-validation framework, we were able to determine the ideal λ value of 0.0173 in the LASSO analysis (Fig. 4A,B) . The genes with non-zero coefficients in the LASSO analysis were then subjected to a RSF study, which led to the identification of a final set of 12 genes (Supplementary file 5) . Furthermore, in both the training and validation sets, patients in the high-risk group had significantly worse OS compared to those in the low-risk group (p < 0.001; Fig. 4D,E,F ). Similarly, the low-risk group had a significantly higher DSS than the high-risk group (p < 0.001; Fig. 4G ). Evaluation of the ARS model The ROC curve analysis shows that ARS has a high discriminative ability. The AUC of the ARS training set are 0.66, 0.66 and 0.66, respectively. Furthermore, at 1, 3 and 5 year intervals, the AUC of the training sets GSE29013 and GSE42127 are 0.93, 0.77, 0.69 and 0.77, 0.72, 0.66 respectively (Fig. 5A, B,C ). Since clinical indicators are often utilized in clinical practice to determine the prognosis of patients with non-small cell lung cancer, we assessed the relationship between ARS and a number of clinical markers. We found significant differences in stage, T, N and M between the high and low risk groups in the TCGA-LUAD dataset (Fig. 5D,G) . Furthermore, we observed (p < 0.001, Wilcox test) that individuals with stage T3-4 had significantly higher risk scores than those with stage T1-2 (Fig. 5E,F) . These results suggest that ARS is associated with a worse outcome in people with NSCLC. It's interesting to note that ARS was able to predict N stage in NSCLC patients (Fig. 5H) . The diagnostic ROC curve indicated that the AUC of ARS for predicting N stage in NSCLC patients was 0.671, indicating its potential use in clinical practice for forecasting the development of metastatic NSCLC (Fig. 5I) . Creation and verification of a nomogram in conjunction with clinical features We conducted univariate and multivariate Cox regression analysis using the TCGA-LUAD dataset to ascertain if ARS is an independent prognostic factor for NSCLC (Fig. 6A) . In univariate analysis, our results showed that ARS was a significant risk factor for OS (HR > 1, p < 0.001). Furthermore, ARS maintained its independent predictive status for OS in multivariate analysis (HR = 1.465, CI = 1.235–1.739, p < 0.001), demonstrating its significant prognostic capacity for patients with NSCLC (Fig. 6B) , thus validating the generalisability of our results across different datasets. In order to increase the clinical applicability of the ARS, we created a column-line diagram based on the ARS and clinical characteristics (Fig. 6E) . The calibration curves showed a strong correlation between the actual measurements and the predictions made using the column-line diagram (Fig. 6D) . Furthermore, the C-index outperformed other clinical variables in predicting OS from 1 to 10 years, demonstrating the stable and superior predictive performance of the column-line diagram (Fig. 6C). DCA showed that column-line plots had a higher net therapeutic benefit compared to other clinical features (Fig. 6F) . These findings imply that ARS-based column-line plots are a dependable and accurate method for personalised prognostic prediction in non-small cell lung cancer. The molecular processes behind the ARS in the bulk transcriptome To explore the molecular pathways behind the correlation between ARS and NSCLC prognosis, we used functional enrichment analysis. Based on the GO gene set, we performed GSEA analysis and found that the high-risk group was enriched for positive regulation of cell proliferation, differentiation and apoptosis processes as well as energy metabolism processes, whereas the low-risk group was enriched for positive regulation of oxidative stress, bile acid and salt metabolism processes as well as characteristic immunity (Fig. 7A) . Furthermore, as shown in Fig. 7B and Fig. 7C , the low-risk group showed higher activity in pathways related to HALLMARK_BILE_ACID_METABOLISM, HALLMARK_HEME_METABOLISM and HALLMARK_FATTY_ACID_METABOLISM, while the high-risk group showed stronger activity in pathways associated with IL6_JAK_STAT3_SIGNALING, HALLMARK_G2M_CHECKPOINT, MYC_TARGETS_V2, HALLMARK_HYPOXIA and HALLMARK_E2F_TARGETS. Correlation studies between ARS and Signature Pathway scores provided further support for these results, indicating a substantial association between ARS and metabolic pathways and biological processes associated with cancer. We used KM curve analysis to determine whether Hallmark pathways were associated with NSCLC prognosis. Poor prognosis was associated with pathways positively associated with ARS, including IL6_JAK_STAT3_SIGNALING ,and HALLMARK_HYPOXIA (Fig. 7E,F) . On the other hand, prognosis was positively correlated with pathways inversely associated with ARS, including HALLMARK_E2F_TARGETS (Fig. 7D,G) . These findings suggest that the different prognostic outcomes seen in ARS risk subgroups may be due to either activation or inhibition of these mechanisms. Landscape of genomic variation and tumour heterogeneity across ARS categories Intra-tumor hetero(ITH) geneity is a well-known genetic feature of cancer that arises from the accumulation of mutations ( 24 ). It has been shown that ITH is linked to a higher risk of cancer and treatment resistance ( 25 ). Higher MATH scores were linked to higher ITH in this study, which assessed ITH in patients with NSCLC using the mutant allele tumour heterogeneity (MATH) approach. The high-risk group of NSCLC patients had higher MATH scores, as shown in Fig. 8A . Patients in the high risk + high MATH group had a considerably poorer prognosis than patients in the low risk + low MATH group when we combined ITH and NSCLC (p < 0.001). This suggests that using these two measures together may improve prognosis for NSCLC patients. (Fig. 8B) . To examine the differences in genomic mutations across the ARS subgroups, we have shown the mutation patterns between the high and low risk groups in Fig. 8C,D . Different aspects of the mutations were observed between the two groups. For example, DNA repair and apoptosis are mediated by the important tumour suppressor gene TP53( 26 ). The high-risk group saw a mutation frequency of 56% as opposed to 45% for the low-risk group. When MUC16, another tumour-promoting factor, is mutated, it can lead to immunological escape, chemotherapy resistance and a poor prognosis. The mutation frequency was 35% in the high-risk group and 40% in the low-risk group( 27 , 28 ). The differences in prognosis seen in the risk sub-groups of the ARS may be explained by the different mutational landscapes between the risk classes. We explored the link between co-occurring and exclusive mutations in the top 20 mutated genes in both the high and low risk categories. According to the results, there was a higher incidence of co-occurring mutations in the high-risk group (Fig. 8E, F) . CNV of the top 11 genes with the largest changes between ARS risk categories was also examined (Fig. 8G) . According to our results, the main changes in FLG, KRAS and CSMD3 were due to CNV gain, whereas the main changes in LRP1B, MUC16 and TP53 were due to CNV loss. The connection between the anti-cancer immune cycle, immunotherapy response and the ARS In order to evaluate the immune infiltration status of NSCLC samples, we used the ESTIMATE technique to calculate the immunological score, stroma score, ESTIMATE score, and tumour purity score for NSCLC risk groups. Figure 9A-C shows that the immunity, ESTIMATE, and tumour purity scores were considerably higher in the high-risk group. We also found that 12 anoikis-related genes had strong correlations with tumour infiltrating immune cells, with monocytes and FCGRT and FBP1 having good correlations and macrophage M2 having positive correlations with CTSH and NPC2 (Fig. 9F) . For the purpose of analyzing the variations in particular immune cell infiltration between the high-risk and low-risk groups, we used the CIBERSORT technique to quantify the number of invasive immune cells in each sample (Fig. 9D) . In addition, we validated the result with the ssGSEA algorithm and showed that the high-risk group had higher levels of CD8 T cells and memory B cells. (Fig. 9E) . Using Spearman’s correlation analysis, we then searched for immune cell types that were significantly associated with ARS and found 12 cell types (p < 0.05) (Fig. 9G) . Using overlapping Venn diagrams, we were finally able to identify seven overlapping TME celltypes :T_cells_CD4_memory_resting,T_cells_CD4_memory_activad, Monocytes, Macrophages_M0, Macrophages_M2, Dendritic cells_resting and Mast cells_resting (Fig. 9H) . These findings highlight the significance of these seven forms of immune cell infiltration for the development and course of NSCLC. The relationship between ARS and immunotherapy response Large immune cell infiltrations alone are not sufficient to characterise immune activation and exhaustion due to the complexity of immunological processes and the environment within the tumour. Nevertheless, a more comprehensive comprehension of the anti-cancer mechanism of immune cells may be acquired by evaluating the efficacy of each phase of the immune cycle in combating cancer. This will provide a better direction for immunotherapy. Figure 10A shows that the ARS risk subgroups differed significantly in phases 2, 4, 5 and 7 of the anti-cancer immune cycle. Phase 7 showed that the high-risk group was more active in killing cancer cells. In addition, "Step 4 - Immune cell trafficking to the tumour" was further developed to investigate the recruitment of different immune cells by the ARS risk subgroups. The results showed that helper T cells, regulatory T cells and immune cells including CD16 + monocytes and CD8 + memory cells were more abundant in the high-risk group (Fig. 10B) . These results suggest that people at higher risk have more anticancer activity in the immune cell function cycle. Previous studies have shown that increased levels of immune checkpoint expression are linked to a more favorable response to ICIs ( 29 – 31 ). We therefore looked at immune checkpoint expression levels across ARS risk categories. The majority of immune checkpoints, including TNFRSF25, TIGIT, CTLA-4, PDCD1 (PD1) and LAG3, were significantly overexpressed in the high-risk group, as shown in Fig. 10C . To confirm our findings, we examined the IPS scores derived from the TCIA database. In four categories: ( 1 ) ips_ctla4_pos_pd1_pos (CTLA4 + /PD1- treatment), ( 2 ) ips_ctla4_pos_pd1_neg (CTLA4 + /PD1- treatment), ( 3 ) ips_ctla4_neg_pd1_pos (CTLA4-/PD1-treatment), and ( 4 ) ips_ctla4_pos_pd1_pos (CTLA4-/PD1-treatment), and ( 4 ) ips_ctla4_neg_pd1_neg (CTLA4/PD1 treatment), higher IPS scores were associated with a better response to ICI treatment, which included PD-1 inhibitor and CTLA4 inhibitor treatment. Our findings indicate that the immune checkpoint inhibitors CTLA4 + /PD1 + and CTLA4 + /PD1 - treatment resulted in a substantially greater IPS in the high-risk groups. This suggests that patients in the high-risk group responded more favourably to both anti-CTLA4 and anti-CTLA4 + /PD1 therapy and that the PD-1 and anti-CTLA4 combination therapy was better compared to the low-risk group (Fig. 10D-G). We included the atezolizumab-treated IMvigor210 cohort to further validate the prognostic power of ARS in predicting patient response to immunotherapy. We determined the cohort's risk score using the ARS model and then divided the patients into high-risk and low-risk groups. The high-risk group had a significant increase in tumour mutational burden (TMB), a well-established indicator of response to immunotherapy (Fig. 10H) . The percentage of complete remission/partial remission (CR/PR) was significantly higher in the high-risk group according to the chi-squared test, while the low-risk group had a higher number of stable disease/progression cases (SD/PD) (Fig. 10I,J) . In addition, those with CR/PR had significantly higher risk scores compared to those with SD/PD (Fig. 10K) . In summary, these results confirm the ARS's capability to forecast the effectiveness of immunotherapy and indicate that patients classified as high-risk will have greater advantages from the medication. Examination of the relationship between medication sensitivity and the ARS First-line treatment for advanced NSCLC often consists of cytotoxic drugs and multi-targeted tyrosine kinase inhibitors (TKIs)( 32 , 33 ). In order to do this, we investigated the sensitivity of the ARS risk subgroup to several cytotoxic drugs (cisplatin and paclitaxel) and tyrosine kinase inhibitors (ositinib, gefitinib and gemcitabine). According to our research, the half IC50 of gefitinib was much lower in the low-risk group, and there was a positive correlation between the risk score and the IC50 of gefitini. On the other hand, ositinib, gemcitabine, cisplatin and paclitaxel had lower IC50s for the high-risk group, and there was a negative correlation between risk scores and IC50s for these drugs (Fig. 11) . The findings indicate that those classified as high-risk shown a more favorable response to the therapeutic regimen consisting of ositinib, gemcitabine, cisplatin, and paclitaxel. Conversely, those classified as low-risk exhibited a higher degree of sensitivity to gefitinib. Discussion In order to determine consistent and trustworthy predictive characteristics for NSCLC, we first developed an exclusive computational framework including 10 advanced machine learning algorithms and 101 potential combinations. The results of our analysis revealed ARS, which outperformed other research attempting to create a prognostic classifier for NSCLC programmed cell death-related prognosis in terms of predictive accuracy and clinical translational importance. Second, we risk stratified NSCLC patients using the ARS and evaluated their response to immunotherapy and sensitivity to first-line NSCLC drugs such as cytotoxic and TKIs. These results provide logical recommendations for the use of immunotherapy and chemotherapy in clinical practice, a significant step towards a more successful personalised medicine strategy. Furthermore, we have discovered potential medications that impede the advancement of NSCLC towards high-risk characteristics, presenting novel insights into preventative strategies for the illness. In addition, to gain a better understanding of ARS, we performed extensive multi-omics analyses, including bulk transcriptome and genome analyses, in contrast to other research that only examined the predictive power of the features. Our studies have identified the underlying processes and molecular underpinnings at different histological stages, providing evidence for the substantial correlation between ARS and NSCLC progression and prognosis. These findings also provide biological rationale and evidence for ARS that may guide personalised treatment strategies. Finally, we used a unique bioinformatics strategy integrating AUCell, UCell, GSVA, Singscore, AddModuleScore, ssGSEA and WGCNA algorithms to discover ARS at the bulk transcriptome level. Using this method, we were able to identify genes associated with ANO that could serve as potential targets for therapeutic intervention in NSCLC specifically. Furthermore, these discoveries offer fresh perspectives for future investigations of ANO in NSCLC. We have identified genes specifically associated with anoikis in NSCLC patients and used these genes to build a predictive model. One of the identified proteins, FBP1, has been identified as a tumour suppressor that is missing in many cancers. FBP1 functions as a protein phosphatase and plays an important role in inhibiting cancer progression( 34 – 36 ). According to a study by Emilie Dalloneau et al, global expression of FCGRT mRNA may indicate the abundance of antigen-presenting cells and the immune response to tumours in NSCLC. This finding could potentially help the decision-making process for patients with NSCLC( 37 ). MT2A is responsible for the production of metallothionein 2A protein. Expression levels of MT2A mRNA are considered an indicator of poor prognosis in lung cancer patients. In addition, they have the ability to suppress MT2A expression, leading to cell death and apoptosis in tumour cells( 38 , 39 ). NUPR1 is a nuclear transcriptional regulator that can be induced to be expressed under various stressful environments and conditions and is involved in a variety of cellular physiological processes. Xiaotong Guo et al. report that silencing NUPR1 by tail vein injection of lentivirus-encoded shRNA also inhibited tumour growth in vivo. In addition, lentivirus-mediated RNAi targeting of NUPR1 significantly decreased the growth of NSCLC cells and triggered apoptosis in vitro ( 40 ). According to the findings of Ji Young Kim et al, ARRB2 was found to be functionally involved in both signalling axes (TRAF6-TAB2 and TRAF6-BECN1 signalling pathways) in response to TLR3/4-stimulated NF-κB activation and autophagy induction to orchestrate lung cancer progression( 41 ). In addition, targeting RGS1 may increase the efficacy of immunotherapy, as it has been shown to be the highest RGS family gene positively associated with immunogenicity. Baojun Wang et al. describe RGS1 as a potential target for immunotherapy( 42 ). HLA-DQA1 is typically expressed on antigen-presenting cells and is part of the human leukocyte antigen (HLA) complex. It is required for leukaemia immunotherapy and has been shown to enhance immune responses( 43 , 44 ). In this study, we emphasize the value of ARS in directing personalized therapy and targeted prevention in NSCLC, which may assist improve patient outcomes and save wasteful treatment expenditures. Overall, ARS may be a useful tool to provide physicians with the essential data for individualised drug selection. However, there are several drawbacks to this study. First, although we assessed and verified ARS in the training and validation datasets, further multicentre prospective investigations are needed to confirm our results. Secondly, further in vivo and in vitro research is needed to clarify the biological roles of ARS in NSCLC. Finally, although we hypothesised how sensitive different small molecule drugs would be to ARS risk categories, further research using in vitro drug testing and clinical trials is needed to confirm our findings. Conclusion In this work, we developed an anoikis-related signature that might be a helpful tool for prognosis prediction, prevention and personalised therapy of NSCLC patients. In addition, we provided new insights from the fields of bulk transcriptomics and genomics into the molecular pathways underlying the initiation and development of NSCLC. Declarations Author Contribution Statement Yuqi Ma : Writing – original draft, Project administration, Methodology, Formal analysis, Data curation, Conceptualization. Jia Li : Writing – original draft Conceptualization, Methodology，Visualization and formal analysis. Tao Shen : Review & editing. Code availability The methods of analysis and the packages used are described in the "Materials and methods" section. All other codes are available from the corresponding author upon request. Data availability The data achieved and analyzed in the current study are available in the TCGA repository (https://portal.gdc.cancer.gov/) and GEO database (https://www. ncbi.nlm.nih.gov/geo/). Finding Not applicable. Declaration of Competing Interest All authors voluntarily participated in this study and declare no competing interests. References Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: a cancer journal for clinicians. 2021;71(3):209–49. Schabath MB, Cote ML. Cancer Progress and Priorities: Lung Cancer. Cancer epidemiology, biomarkers & prevention: a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 2019;28(10):1563–79. Garon EB, Hellmann MD, Rizvi NA, Carcereny E, Leighl NB, Ahn MJ, et al. Five-Year Overall Survival for Patients With Advanced Non–Small-Cell Lung Cancer Treated With Pembrolizumab: Results From the Phase I KEYNOTE-001 Study. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 2019;37(28):2518–27. Gout DY, Groen LS, van Egmond M. The present and future of immunocytokines for cancer treatment. Cellular and molecular life sciences: CMLS. 2022;79(10):509. Ribas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade. Science (New York, NY). 2018;359(6382):1350–5. Heeke S, Hofman P. Tumor mutational burden assessment as a predictive biomarker for immunotherapy in lung cancer patients: getting ready for prime-time or not? Translational lung cancer research. 2018;7(6):631–8. Zhang D, Zhou X, Zhang K, Yu Y, Cui SW, Nie S. Glucomannan from Aloe vera gel maintains intestinal barrier integrity via mitigating anoikis mediated by Nrf2-mitochondria axis. International journal of biological macromolecules. 2023;235:123803. Zhong X, Rescorla FJ. Cell surface adhesion molecules and adhesion-initiated signaling: understanding of anoikis resistance mechanisms and therapeutic opportunities. Cellular signalling. 2012;24(2):393–401. Ye G, Yang Q, Lei X, Zhu X, Li F, He J, et al. Nuclear MYH9-induced CTNNB1 transcription, targeted by staurosporin, promotes gastric cancer cell anoikis resistance and metastasis. Theranostics. 2020;10(17):7545–60. Buchheit CL, Angarola BL, Steiner A, Weigel KJ, Schafer ZT. Anoikis evasion in inflammatory breast cancer cells is mediated by Bim-EL sequestration. Cell death and differentiation. 2015;22(8):1275–86. Chi H, Jiang P, Xu K, Zhao Y, Song B, Peng G, et al. A novel anoikis-related gene signature predicts prognosis in patients with head and neck squamous cell carcinoma and reveals immune infiltration. Frontiers in genetics. 2022;13:984273. Sun Z, Zhao Y, Wei Y, Ding X, Tan C, Wang C. Identification and validation of an anoikis-associated gene signature to predict clinical character, stemness, IDH mutation, and immune filtration in glioblastoma. Frontiers in immunology. 2022;13:939523. Mariathasan S, Turley SJ, Nickles D, Castiglioni A, Yuen K, Wang Y, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature. 2018;554(7693):544–8. Cai Z, Zhou F. A novel Anoikis and immune-related genes marked prognostic signature for colorectal cancer. Medicine. 2022;101(46):e31127. Diao X, Guo C, Li S. Identification of a novel anoikis-related gene signature to predict prognosis and tumor microenvironment in lung adenocarcinoma. Thoracic cancer. 2023;14(3):320–30. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC bioinformatics. 2013;14:7. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research. 2015;43(7):e47. Ma D, Jiang YZ, Liu XY, Liu YR, Shao ZM. Clinical and molecular relevance of mutant-allele tumor heterogeneity in breast cancer. Breast cancer research and treatment. 2017;162(1):39–48. Mroz EA, Tward AD, Hammon RJ, Ren Y, Rocco JW. Intra-tumor genetic heterogeneity and mortality in head and neck cancer: analysis of data from the Cancer Genome Atlas. PLoS medicine. 2015;12(2):e1001786. Mroz EA, Rocco JW. MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma. Oral oncology. 2013;49(3):211–5. Rajput A, Bocklage T, Greenbaum A, Lee JH, Ness SA. Mutant-Allele Tumor Heterogeneity Scores Correlate With Risk of Metastases in Colon Cancer. Clinical colorectal cancer. 2017;16(3):e165-e70. Geeleher P, Cox N, Huang RS. pRRophetic: an R package for prediction of clinical chemotherapeutic response from tumor gene expression levels. PloS one. 2014;9(9):e107468. Yang C, Zhang H, Chen M, Wang S, Qian R, Zhang L, et al. A survey of optimal strategy for signature-based drug repositioning and an application to liver cancer. eLife. 2022;11. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Jr., Kinzler KW. Cancer genome landscapes. Science (New York, NY). 2013;338(6127):1546–58. Dagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nature reviews Clinical oncology. 2018;15(2):81–94. Chang J, Zhao X, Wang Y, Liu T, Zhong C, Lao Y, et al. Genomic alterations driving precancerous to cancerous lesions in esophageal cancer development. Cancer cell. 2023;41(12):2038-50.e5. Zhang XY, Hong LL, Ling ZQ. MUC16: clinical targets with great potential. Clinical and experimental medicine. 2024;24(1):101. Shah A, Chaudhary S, Lakshmanan I, Aithal A, Kisling SG, Sorrell C, et al. Chimeric antibody targeting unique epitope on onco-mucin16 reduces tumor burden in pancreatic and lung malignancies. NPJ precision oncology. 2023;7(1):74. Xu-Monette ZY, Zhou J, Young KH. PD-1 expression and clinical PD-1 blockade in B-cell lymphomas. Blood. 2018;131(1):68–83. Qin S, Xu L, Yi M, Yu S, Wu K, Luo S. Novel immune checkpoint targets: moving beyond PD-1 and CTLA-4. Molecular cancer. 2019;18(1):155. Chauvin JM, Zarour HM. TIGIT in cancer immunotherapy. Journal for immunotherapy of cancer. 2020;8(2). Jänne PA, Baik C, Su WC, Johnson ML, Hayashi H, Nishio M, et al. Efficacy and Safety of Patritumab Deruxtecan (HER3-DXd) in EGFR Inhibitor-Resistant, EGFR-Mutated Non-Small Cell Lung Cancer. Cancer discovery. 2022;12(1):74–89. Jiang L, Liu J. Immunological effect of tyrosine kinase inhibitors on the tumor immune environment in non-small cell lung cancer. Oncology letters. 2022;23(5):165. Zhu W, Chu H, Zhang Y, Luo T, Yu H, Zhu H, et al. Fructose-1,6-bisphosphatase 1 dephosphorylates IκBα and suppresses colorectal tumorigenesis. Cell research. 2023;33(3):245–57. Wang Z, Li M, Jiang H, Luo S, Shao F, Xia Y, et al. Fructose-1,6-bisphosphatase 1 functions as a protein phosphatase to dephosphorylate histone H3 and suppresses PPARα-regulated gene transcription and tumour growth. Nature cell biology. 2022;24(11):1655–65. Zhang G, Tao J, Lin L, Qiu W, Lu Z. Repurposing FBP1: dephosphorylating IκBα to suppress NFκB. Cell research. 2023;33(6):419–20. Dalloneau E, Baroukh N, Mavridis K, Maillet A, Gueugnon F, Courty Y, et al. Downregulation of the neonatal Fc receptor expression in non-small cell lung cancer tissue is associated with a poor prognosis. Oncotarget. 2016;7(34):54415–29. Werynska B, Pula B, Muszczynska-Bernhard B, Gomulkiewicz A, Piotrowska A, Prus R, et al. Metallothionein 1F and 2A overexpression predicts poor outcome of non-small cell lung cancer patients. Experimental and molecular pathology. 2013;94(1):301–8. Yamasaki M, Nomura T, Sato F, Mimata H. Metallothionein is up-regulated under hypoxia and promotes the survival of human prostate cancer cells. Oncology reports. 2007;18(5):1145–53. Kim YJ, Hong SE, Jang SK, Park KS, Kim CH, Park IC, et al. Knockdown of NUPR1 Enhances the Sensitivity of Non-small-cell Lung Cancer Cells to Metformin by AKT Inhibition. Anticancer research. 2022;42(7):3475–81. Kim JY, Shin JH, Kim MJ, Kang Y, Lee JS, Son J, et al. β-arrestin 2 negatively regulates lung cancer progression by inhibiting the TRAF6 signaling axis for NF-κB activation and autophagy induced by TLR3 and TLR4. Cell death & disease. 2023;14(7):422. Wang B, Jiang B, Du L, Chen W, Zhang Q, Chen W, et al. Tumor-intrinsic RGS1 potentiates checkpoint blockade response via ATF3-IFNGR1 axis. Oncoimmunology. 2023;12(1):2279800. Cheng L, Liu Q, Wang M, Gu Y, Wang J, Wei Q, et al. Genetic variants in the human leukocyte antigen region and survival of Chinese patients with non-small cell lung carcinoma. Carcinogenesis. 2020;41(9):1203–12. Herr W, Eichinger Y, Beshay J, Bloetz A, Vatter S, Mirbeth C, et al. HLA-DPB1 mismatch alleles represent powerful leukemia rejection antigens in CD4 T-cell immunotherapy after allogeneic stem-cell transplantation. Leukemia. 2017;31(2):434–45. Additional Declarations No competing interests reported. Supplementary Files Supplementaryfile1.docx Supplementaryfile2.pdf Supplementaryfile3.pdf Supplementaryfile4.pdf Supplementaryfile5.pdf Supplementaryfile6.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4640324","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":328604145,"identity":"e0e7542b-5ee0-4573-97db-ba27d0b49df5","order_by":0,"name":"Yuqi Ma","email":"","orcid":"","institution":"Chengdu University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yuqi","middleName":"","lastName":"Ma","suffix":""},{"id":328604147,"identity":"23706cdd-8c42-4d82-bfe4-3eaf266f86e3","order_by":1,"name":"Jia Li","email":"","orcid":"","institution":"Chengdu University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jia","middleName":"","lastName":"Li","suffix":""},{"id":328604149,"identity":"dcd36e2d-2b0b-4a2f-b1de-da49c43cf2c8","order_by":2,"name":"Tao Shen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIiWNgGAWjYNACAxsIzQPlSxChJQ2imofBgFgtDIdJ0KLbfsZM4k3BeXt7iQTGD2/b/sjrNjAfvM3DYJeHS4vZmRwzyTkGtxN7JBKYJee2GRhuO8CWbM3DkFyMU8uBHDNpHoPbCTwSCWzMvG0GCWYHeIAiDAcSG3BpOf8GpOWcPZIW/m/4tdwA23KAsQfJFjYCWp4VW84xSE7sOfOwWXLOOWPDbYfZjMEiuB2WvPHGmz929uztyQc/vCmTkzc73vzwxpsKO5xaGBg4DKCRzghVwwwiDHCqBwL2B/B0MgpGwSgYBaMAKwAAmIZOLQn+H5kAAAAASUVORK5CYII=","orcid":"","institution":"Chengdu University of Traditional Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Tao","middleName":"","lastName":"Shen","suffix":""}],"badges":[],"createdAt":"2024-06-26 06:22:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4640324/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4640324/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":60900985,"identity":"439b00a4-facd-4a19-aee2-081a0fdd2bb0","added_by":"auto","created_at":"2024-07-23 10:52:39","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":808424,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/73e4cf67944a535e2f7a917c.jpg"},{"id":60901000,"identity":"c76b2694-55f1-4938-b381-2dfdf14dbc24","added_by":"auto","created_at":"2024-07-23 10:52:41","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2799138,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/03f36aca819e2285137dd037.jpg"},{"id":60901518,"identity":"e8fefb01-14cf-4b7d-919e-16aebf5c476c","added_by":"auto","created_at":"2024-07-23 11:00:39","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4582478,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/a4ce105f3b6ddfddf696c2d0.jpg"},{"id":60902131,"identity":"d10127ba-b0f1-475f-a55f-384456bb8561","added_by":"auto","created_at":"2024-07-23 11:08:40","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":6367157,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/56657dd38b8fd063c3074e16.jpg"},{"id":60900989,"identity":"543a04b0-041b-4e7e-bb8a-4f1b54587303","added_by":"auto","created_at":"2024-07-23 10:52:40","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":5855466,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/0a788830fe109a944cf1e364.jpg"},{"id":60900987,"identity":"3c03000d-6eca-4a34-b944-2183ec5ba627","added_by":"auto","created_at":"2024-07-23 10:52:40","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2091738,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/16a4dd59a65ad979f3e2f0de.jpg"},{"id":60901519,"identity":"488bc2de-cac7-4d44-ada5-cf30553a1a2a","added_by":"auto","created_at":"2024-07-23 11:00:40","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":2460343,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/055c949ebfedeb1877743270.jpg"},{"id":60900993,"identity":"c8f2dde8-4bd6-4784-8702-5796cae4d856","added_by":"auto","created_at":"2024-07-23 10:52:40","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":5059318,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/17463dcabe165e535b1f688c.jpg"},{"id":60902132,"identity":"0115a9d9-ad70-4250-a0a8-92059379132b","added_by":"auto","created_at":"2024-07-23 11:08:40","extension":"jpg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":4092298,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure9.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/993e1ef36f070f0f6d3ff179.jpg"},{"id":60900998,"identity":"11b49922-3dcb-4fc3-b5b8-d110c412e8c3","added_by":"auto","created_at":"2024-07-23 10:52:41","extension":"jpg","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":3726772,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure10.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/f802d61cb4a073076d0dfa4a.jpg"},{"id":60900999,"identity":"7eb82763-d7b1-4273-bba1-f0a61309a797","added_by":"auto","created_at":"2024-07-23 10:52:41","extension":"jpg","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":968316,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"figure11.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/57c679af4b751a42b6e2aab4.jpg"},{"id":61132230,"identity":"391d5bd5-9dee-4ea3-8039-1dd59000a29f","added_by":"auto","created_at":"2024-07-26 04:34:56","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":39633255,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/53e6dbfb-7c26-487f-9001-325ed09734b9.pdf"},{"id":60900992,"identity":"7296e5a0-7fbe-4bc4-b958-d73cf1b83efc","added_by":"auto","created_at":"2024-07-23 10:52:40","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":18455,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/557e4d752c2fd7e3dc6476ef.docx"},{"id":60901003,"identity":"b079252a-c92a-4c50-8c28-44c88f6a63df","added_by":"auto","created_at":"2024-07-23 10:52:42","extension":"pdf","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":11660784,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/3f60ff4bf59201f9a4884f14.pdf"},{"id":60901521,"identity":"2ffa3052-3edd-442f-95a8-8a304e027ce8","added_by":"auto","created_at":"2024-07-23 11:00:40","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":5078,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/f47d970875f6ddf1c8f95ff9.pdf"},{"id":60901005,"identity":"7d2e9776-beb7-4478-a855-2146d2fcbb2c","added_by":"auto","created_at":"2024-07-23 10:52:42","extension":"pdf","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":1466946,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/fef6e9f75c427421ca82d39b.pdf"},{"id":60900996,"identity":"d75d604d-d005-496b-ac4a-1171d8e26c22","added_by":"auto","created_at":"2024-07-23 10:52:40","extension":"pdf","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":6422,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/ac2bda4f2f22138373fe50d3.pdf"},{"id":60901523,"identity":"905daae5-612a-4745-91c2-6572d9b51110","added_by":"auto","created_at":"2024-07-23 11:00:40","extension":"docx","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":16076,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile6.docx","url":"https://assets-eu.researchsquare.com/files/rs-4640324/v1/e0250876186e8b902f6fb88b.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-omics analysis reveals an anoikis-related signature for non-small cell lung cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAccounting for 18% of all cancer deaths worldwide, lung cancer is the leading cause of cancer deaths(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Non-small cell lung cancer (NSCLC), with an estimated 5-year survival rate of 10\u0026ndash;23%, is the most common type of lung cancer(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). NSCLC is typically diagnosed at a late stage, which is a major limitation to the effectiveness of treatment(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). As a result, there is an urgent need for innovative biomarkers to help predict the prognosis of patients with NSCLC. These biomarkers would allow timely clinical intervention to slow disease progression.\u003c/p\u003e \u003cp\u003eImmune checkpoint inhibitors (ICIs) have been generally accepted by patients and have changed the management of advanced NSCLC. ICIs stimulate the body's immune system to detect and remove cancer cells(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). However, these immune-based therapies only assist a fraction of NSCLC patients because primary resistance remains prominent in many situations (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn recent years, extensive studies of various cell death modalities have identified cell death as a critical mechanism that can be exploited in therapy. Anoikis is a form of programmed cell death that induces cell death primarily by disrupting endogenous mitochondrial pathways or activating exogenous pathways through cell surface death receptors(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Therefore, the identification of biomarkers or characteristics that are associated with prognosis and response to ICIs in patients with NSCLC is essential.\u003c/p\u003e \u003cp\u003eRecently, a growing number of anoikis-related signature (ARS) have been identified in various cancers. Ye et al. showed that MYH9 expression is upregulated in metastatic gastric cancer tissue and is associated with poor prognosis in gastric cancer patients. Further studies showed that MYH9 binds to the CTNNB1 promoter through its DNA binding domain, conferring resistance to anoikis in tumour cells(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). GDH1-mediated reprogramming of glutamine catabolism has been shown to be a driver of anoikis resistance and tumour metastasis in LKB1-deficient lung cancer(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). In addition, ARS have been shown to be associated with prognosis in several tumours(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). However, comprehensive analyses of the impact of anoikis in NSCLC remain scarce. For instance, the use of biomarkers associated with anoikis to stratify patients according to prognosis and to predict their response to immunotherapy and chemotherapy. This is an area of significant importance for future studies.\u003c/p\u003e \u003cp\u003eThis research intended to examine the features of ARS at various histological levels. We turned both single-cell and bulk transcriptomes to find ARS. To generate consistent ARS, we deployed a unique machine learning framework that consisted of 10 machine learning methods and 101 combinations. In order to improve the implementation of ARS, we conducted an assessment to see whether ARS could accurately forecast the incidence, progression, and spread of NSCLC. In addition, we developed a nomogram that incorporates ARS to offer a precise and statistical approach for estimating patient prognosis in clinical settings. We also analysed how different at-risk subgroups responded to treatment with ICIs and their sensitivity to first-line NSCLC therapies such as tyrosine kinase inhibitors (TKIs) and cytotoxic drugs. Our goal is to personalize prognostic prediction and treatment for NSCLC patients. See the flow chart \u003cb\u003e(Fig.\u0026nbsp;1)\u003c/b\u003e for details.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eData collection and processing\u003c/h2\u003e \u003cp\u003eWe obtained RNA-seq data and accompanying clinical data for NSCLC from The Cancer Genome Atlas (TCGA) and isolated transcripts per million (TPM) values for further analysis. Samples with incomplete or missing clinicopathological information were excluded, resulting in a dataset of 501 TCGA patients. In addition, we utilized the NSCLC single-cell RNA sequencing dataset GSE131907 from the Gene Expression Omnibus (GEO) collection. The Mutation Annotation Format (MAF) was used to extract somatic mutation data from TCGA. Copy number variation (CNV) data for TCGA-LUAD patients were collected from the UCSC Xena database.\u003c/p\u003e \u003cp\u003eTo assess the value of prognostic features for forecasting patient outcome following immunotherapy, we included the IMvigor210 cohort. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). In addition, to examine the prediction ability of these indicators for NSCLC incidence, we added two datasets from the GEO database (GSE29013 and GSE42127). To detect ARS, we gathered 338 these genes from previous research.(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). \u003cb\u003e(Supplementary file 1)\u003c/b\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003escRNA analysis\u003c/h2\u003e \u003cp\u003eWe obtained scRNA sequencing data from 44 patients (58 samples) from the GSE131907. We analysed the single-cell sequencing data using the Seurat package. Quality control (QC) process included selecting cells with mitochondrial gene content below 20% and genes that were expressed in no less than three cells within a range of 500 to 7000 in terms of expression levels. Highly variable genes were identified for subsequent analysis with a variable gene number of 2000. The Harmony package suite was used to remove batch effects in the data from the 58 samples. The FindClusters and FindNeighbors routines were used to generate cell clusters, which were then visualized using the tSNE approach. Ultimately, cellular annotation was conducted using marker genes specific to various cell types.\u003c/p\u003e \u003cp\u003eThe expression levels of anoikis (ANO) in all cells were determined using the five techniques of AUCell, UCell, GSVA, Singscore and Addmodulescore. The total score for all cell types was then determined to quantify the activity of ANO in all cell types. Among the eight cell types, we found that epithelial cells had significantly increased ANO activity. We divided the cells into high and low ANO groups. We then looked for genes that were differentially expressed in the two groups so that we could study them further.\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003essGSEA\u003c/h2\u003e \u003cp\u003eSingle-sample gene set enrichment analysis (ssGSEA) has become a commonly used technique for measuring the enrichment scores of certain gene sets in individual samples. The ssGSEA score of each sample reflects the degree of systematic up- or down-regulation of a specific gene set in the sample. We calculated the ANO score for each TCGA-LUAD sample by using ssGSEA in the R package \"GSVA\" in our study(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). In order to determine the possible pathways related to the attributes, we conducted gene set enrichment analysis (GSVA) to compute scores for 50 signature pathways. We next analyzed the pathways that showed significant differences between the high-risk and low-risk groups using the 'limma' package(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Furthermore, to clarify the biological processes (BP), cellular components (CC), and molecular functions (MF) associated with the various risk subgroups, we used the R package 'clusterProfiler' with a false discovery rate (FDR) threshold of less than 0.25 and an absolute normalized enrichment score (NES) greater than 1.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eWGCNA\u003c/h2\u003e \u003cp\u003eWGCNA is a methodical biological technique that examines patterns of genetic linkages among various samples and discovers genomes that are highly synergistically changed. We used the WGCNA package to conduct a WGCNA analysis using the TCGA-LUAD dataset in our research. First, an acceptable soft threshold β was established to meet the conditions for designing a scale-free network. A dynamic tree-cutting method was applied for gene clustering and module identification. Finally, the module having the strongest connection with ANO scores were picked for further analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eConstructing prognostic models by integrating several machine learning approaches\u003c/h2\u003e \u003cp\u003eWe utilized the limma package to undertake a difference analysis between normal and tumour samples in the TCGA-LUAD dataset. Subsequently, we performed bulk RNA-seq sequencing of differentially expressed genes (DEGs) that were identified by WGCNA inside modules associated with anoikis. We designated these sequenced genes as ARS because they were shown to be associated with anoikis in both the bulk and single-cell transcriptome analyses. In order to create resilient prognostic models with exceptional predicted precision, we adhered to the following procedures:\u003c/p\u003e \u003cp\u003ea: Initially, we used univariate Cox regression analysis to identify ARS that may have predictive significance in the TCGA-LUAD dataset.\u003c/p\u003e \u003cp\u003eb: The TCGA-LUAD dataset was split into two sets at random, one for internal validation and the other for training, in order to guarantee a fair distribution of clinical characteristics across the two sets. There are ten machine learning algorithms: GBM, Lasso, Ridge, Stepwise Cox, CoxBoost, Random Survival Forest (RSF), Elastic Net (Enet), Supervised Principal Component (SuperPC), Partial Least Squares Regression Cox (plsRcox), and Survival Support Vector Machine (Survival-SVM). Using a tenfold cross-validation framework, we organized 101 combinations of these 10 techniques for variable selection and model creation in the TCGA-LUAD training dataset.\u003c/p\u003e \u003cp\u003ec: Two groups were randomly created from the TCGA-LUAD dataset: one for internal validation and one for training. The following 10 machine learning techniques were available: Enet, SuperPC, plsRcox, RSF, Lasso, Ridge, Stepwise Cox, CoxBoost, GBM, and SVM. We selected variables and built models for 101 combinations of these 10 approaches in the TCGA-LUAD training dataset using a tenfold cross-validation framework.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003ePredictive nomogram formation and survival analysis\u003c/h2\u003e \u003cp\u003eBased on the median ARS risk score, the TCGA training set was split into high-risk and low-risk groups. Then, using the R package \"survminer,\" we did KM curve analysis to see whether the high- and low-risk groups' OS and DSS differed significantly from one another (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In addition, we evaluated the sensitivity and specificity of ARS in predicting OS in NSCLC patients by doing ROC curve analysis using the \"timeROC\" software program. We also contrasted other clinical features with AUC of ARS. Furthermore, we investigated the relationship between ARS and a number of clinical attributes, such as grade, age, gender, stage, T, M, and N. Using TCGA-LUAD, we ran univariate and multivariate Cox regression analysis to see whether ARS was a reliable predictor of survival for patients with non-small cell lung cancer. Our goal was to increase the prediction capacity and prognostic accuracy of our model. To do this, we created a nomogram that quantified the predicted survival of NSCLC patients by combining clinical characteristics and ARS. Finally, we assessed the precision and accuracy of the nomogram using ROC curves, C-indices, and calibration curves, as well as its net clinical benefit using decision curve analysis (DCA).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of genomic variation between ARS risk subgroups\u003c/h2\u003e \u003cp\u003eMutant Allele Tumor Heterogeneity (MATH) is a method of quantifying ITH based on the distribution of mutant alleles. MATH scores provide a quantifiable and measurable assessment of ITH(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). Research has investigated the predictive relevance of MATH in a range of tumour types(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e), including breast, colorectal and head and neck malignancies(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). For this investigation, we calculated each NSCLC patient's MATH score using a previously published method. Subsequently, we conducted survival analyses by using their MATH score.\u003c/p\u003e \u003cp\u003eIn order to look at somatic mutations linked to ARS, we created waterfall plots using the maftools package, which show the mutational landscape of NSCLC patients in high- and low-risk groups. We also performed copy number variation (CNV) analysis on 30 genes that showed the most variation between the high and low-risk groups.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eAn extensive perspective of immunological characteristics and responsiveness to immunological checkpoint inhibitor therapy\u003c/h2\u003e \u003cp\u003eOne important component of tumor immunotherapy is the anti-cancer immune cycle, which consists of seven stages: releasing the cancer antigen (step1), presenting the cancer antigen (step2), triggering and activation (step3), transporting immune cells to the tumour (step4), infiltrating the tumour (step5), having T-cells recognize the tumour (step6), and killing the tumour (step7). We quantified the infiltration of 22 immune cells using the CIBERSORT method in order to examine the connection between ARS and immune cell infiltration in the NSCLC tumour microenvironment. We also used the ESTIMATE and ssGSEA to verify the accuracy of the CIBERSORT findings.\u003c/p\u003e \u003cp\u003eWe used the Immunophenotype Scoring (IPS) system to predict immunogenicity based on immunomodulators, immunosuppressive cells, MHC molecules, and effector cells. The IPS system uses a machine learning approach to build IPS scores based on unbiased gene expression of representative cell types. Higher IPS scores are indicative with a better immunotherapy response. The IPS ratings for patient samples from TCGA-LUAD were obtained using the Cancer Immunome Atlas (TCIA) database. (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://tcia.at/at-home\u003c/span\u003e\u003cspan address=\"https://tcia.at/at-home\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eImportance of the ARS for medication sensitivity\u003c/h2\u003e \u003cp\u003eTo individualise treatment, we predicted chemotherapy sensitivity in colorectal cancer patients with different ARS risk scores using the R package pRRophetic(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). By employing pRRophetic to match patient tissue gene expression patterns to cancer cell line patterns, half-maximal inhibitory concentrations (IC50) were found. The difference in medication IC50 between the high and low risk groups was investigated using the Wilcoxon test. When p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, it was deemed statistically significant.\u003c/p\u003e \u003cp\u003eWe also used the Connection Map of Gene Expression Profiles (CMap) database to evaluate substances that might be activating or inhibiting agents. According to Yang et al(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). web-based studies of CMap databases are less likely to yield relevant results than CMap analyses using the R-based \"extreme sum\" (XSum) technique. We used the XSum method to determine CMap scores. If the drug has a CMap score between \u0026minus;\u0026thinsp;1 and 0, it may be able to help people with ccRCC by acting as a cancer drug.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eR software (versions: R 4.3.2) was used for all statistical analyses. The chi-squared test was used to analyse differences in clinical characteristics between the internal validation and training sets. The Wilcoxon test was a non-parametric method for estimating the difference between two non-normally distributed variables. Significant changes in DEGs were evaluated using FDR-corrected p-values. To investigate independent prognostic factors, R was used to conduct both univariate and multivariate Cox regression analysis. ROC curve analysis and AUC were computed using the \"timeROC\" package to evaluate the model's efficacy. The association between immune cell infiltration and risk score was examined using Spearman's correlation analysis.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eCharacteristics of anoikis in single cell transcriptome\u003c/h2\u003e \u003cp\u003eWe acquired 208,506 cells from 44 NSCLC patients' single-cell RNA sequencing data. We successfully integrated 58 samples using the Harmony software program in order to remove batch effects, as shown in \u003cb\u003eFig.\u0026nbsp;2A\u003c/b\u003e. The top 2000 variant genes were then subjected to downscaling using principal component analysis (PCA) and t-distributed stochastic neighbour embedding (t-SNE). After that, at a resolution of 1.5, clustering was done to combine all of the cells into 54 clusters \u003cb\u003e(Supplementary file 2)\u003c/b\u003e. We classified the cells into eight main clusters, NK cell, B lymphocytes, T lymphocytes, Epithelial cell, Fibroblasts, Endothelial cell, Myeloid cell, Mast cell using flag genes specific to various celltypes. The top 5 marker genes for each cell group are shown on the heatmap \u003cb\u003e(Fig.\u0026nbsp;2B)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eThe expression levels of anoikis (ANO) in all cells were determined using the five techniques of AUCell, UCell, GSVA, Singscore and Addmodulescore \u003cb\u003e(Fig.\u0026nbsp;2D)\u003c/b\u003e. The total score for all cell types was then determined to quantify the activity of ANO in all cell types \u003cb\u003e(Fig.\u0026nbsp;2C)\u003c/b\u003e. Of the eight cell types, we found a significant increase in ANO activity in epithelial cells, followed by fibroblast cells. Based on ANO activity, we divided the cells into high and low ANO groups. We then looked for genes that were differentially expressed in the two groups so that we could study them further. Bulk RNA sequencing to identify genes and the hub module associated with ANO.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003essGSEA\u003c/h2\u003e \u003cp\u003eWhen evaluating alterations in biological processes and pathway activity in particular samples, the ssGSEA approach is often used. For each TCGA-LUAD sample in this investigation, we calculated ANO activity scores using the ssGSEA method. These scores served as phenotypic data for further WGCNA analysis. We used WGCNA analysis on the TCGA-LUAD dataset to find modules that were significantly associated with ANO scores. More specifically, co-expression networks were constructed by identifying ARS at the single-cell sequencing level after eliminating outlier data \u003cb\u003e(Fig.\u0026nbsp;3A)\u003c/b\u003e. To ensure scale-free topological networks, power\u0026thinsp;=\u0026thinsp;5 was chosen as the ideal soft threshold \u003cb\u003e(Supplementary file 3)\u003c/b\u003e. By adjusting MEDissThres to 0.25 and the minimum module gene number to 60, a total of four modules were found \u003cb\u003e(Fig.\u0026nbsp;3B)\u003c/b\u003e. The brown module and the ANO score in batch RNA-seq showed a good correlation according to our data \u003cb\u003e(cor\u0026thinsp;=\u0026thinsp;0.61, Fig.\u0026nbsp;3C).\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe TCGA-LUAD bulk RNA-seq volcano plot showed DEGs between tumour and normal tissues (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) \u003cb\u003e(Fig.\u0026nbsp;3D)\u003c/b\u003e. After using DEG from bulk RNA-seq to sequence 229 genes in the brown module, we were able to identify a total of 103 genes \u003cb\u003e(Fig.\u0026nbsp;3E)\u003c/b\u003e. These genes were termed ARS, and it is thought that both the whole transcriptome and the transcriptome of a particular cell are affected by ANO. Biological processes (BPs) were found to be significantly enriched in the GO enrichment analysis of ARS \u003cb\u003e(Fig.\u0026nbsp;3F)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eIn addition, we conducted an analysis on the frequency of CNV for the 36 genes. Among them, GMFG exhibited a copy number increase with a frequency above 10%, as seen in \u003cb\u003eFig.\u0026nbsp;3G\u003c/b\u003e. The protein-protein interaction (PPI) network of these genes is shown in \u003cb\u003eSupplementary file 4\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003ePrognostic feature construction with integrated machine learning\u003c/h2\u003e \u003cp\u003eUsing a mixture of 101 machine learning approaches, we analyzed 31 prognostic genes obtained from univariate Cox regression analysis in order to provide consistent ARS. As shown in \u003cb\u003eFig.\u0026nbsp;4H\u003c/b\u003e, using a tenfold cross-validation framework, we fitted 101 prediction models to the training set and determined C-indexes for both the training and validation sets.\u003c/p\u003e \u003cp\u003eThe prediction models with the greatest average C-index among all the models were constructed using the RSF and Lasso algorithms after the analysis of 101 models. Within the training, internal, and external validation sets, this model demonstrated strong prediction performance. After careful screening, we discovered that Lasso\u0026thinsp;+\u0026thinsp;RSF is a prediction model with outstanding accuracy and translational relevance.\u003c/p\u003e \u003cp\u003eBy minimising the biased likelihood and using a tenfold cross-validation framework, we were able to determine the ideal λ value of 0.0173 in the LASSO analysis \u003cb\u003e(Fig.\u0026nbsp;4A,B)\u003c/b\u003e. The genes with non-zero coefficients in the LASSO analysis were then subjected to a RSF study, which led to the identification of a final set of 12 genes \u003cb\u003e(Supplementary file 5)\u003c/b\u003e. Furthermore, in both the training and validation sets, patients in the high-risk group had significantly worse OS compared to those in the low-risk group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; \u003cb\u003eFig.\u0026nbsp;4D,E,F\u003c/b\u003e). Similarly, the low-risk group had a significantly higher DSS than the high-risk group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; \u003cb\u003eFig.\u0026nbsp;4G\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eEvaluation of the ARS model\u003c/h2\u003e \u003cp\u003eThe ROC curve analysis shows that ARS has a high discriminative ability. The AUC of the ARS training set are 0.66, 0.66 and 0.66, respectively. Furthermore, at 1, 3 and 5 year intervals, the AUC of the training sets GSE29013 and GSE42127 are 0.93, 0.77, 0.69 and 0.77, 0.72, 0.66 respectively (Fig.\u0026nbsp;5A,\u003cb\u003eB,C\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eSince clinical indicators are often utilized in clinical practice to determine the prognosis of patients with non-small cell lung cancer, we assessed the relationship between ARS and a number of clinical markers. We found significant differences in stage, T, N and M between the high and low risk groups in the TCGA-LUAD dataset \u003cb\u003e(Fig.\u0026nbsp;5D,G)\u003c/b\u003e. Furthermore, we observed (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Wilcox test) that individuals with stage T3-4 had significantly higher risk scores than those with stage T1-2 \u003cb\u003e(Fig.\u0026nbsp;5E,F)\u003c/b\u003e. These results suggest that ARS is associated with a worse outcome in people with NSCLC. It's interesting to note that ARS was able to predict N stage in NSCLC patients\u003cb\u003e(Fig.\u0026nbsp;5H)\u003c/b\u003e. The diagnostic ROC curve indicated that the AUC of ARS for predicting N stage in NSCLC patients was 0.671, indicating its potential use in clinical practice for forecasting the development of metastatic NSCLC \u003cb\u003e(Fig.\u0026nbsp;5I)\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eCreation and verification of a nomogram in conjunction with clinical features\u003c/h2\u003e \u003cp\u003eWe conducted univariate and multivariate Cox regression analysis using the TCGA-LUAD dataset to ascertain if ARS is an independent prognostic factor for NSCLC \u003cb\u003e(Fig.\u0026nbsp;6A)\u003c/b\u003e. In univariate analysis, our results showed that ARS was a significant risk factor for OS (HR\u0026thinsp;\u0026gt;\u0026thinsp;1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Furthermore, ARS maintained its independent predictive status for OS in multivariate analysis (HR\u0026thinsp;=\u0026thinsp;1.465, CI\u0026thinsp;=\u0026thinsp;1.235\u0026ndash;1.739, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), demonstrating its significant prognostic capacity for patients with NSCLC \u003cb\u003e(Fig.\u0026nbsp;6B)\u003c/b\u003e, thus validating the generalisability of our results across different datasets.\u003c/p\u003e \u003cp\u003eIn order to increase the clinical applicability of the ARS, we created a column-line diagram based on the ARS and clinical characteristics \u003cb\u003e(Fig.\u0026nbsp;6E)\u003c/b\u003e. The calibration curves showed a strong correlation between the actual measurements and the predictions made using the column-line diagram \u003cb\u003e(Fig.\u0026nbsp;6D)\u003c/b\u003e. Furthermore, the C-index outperformed other clinical variables in predicting OS from 1 to 10 years, demonstrating the stable and superior predictive performance of the column-line diagram \u003cb\u003e(Fig.\u0026nbsp;6C).\u003c/b\u003e DCA showed that column-line plots had a higher net therapeutic benefit compared to other clinical features \u003cb\u003e(Fig.\u0026nbsp;6F)\u003c/b\u003e. These findings imply that ARS-based column-line plots are a dependable and accurate method for personalised prognostic prediction in non-small cell lung cancer.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eThe molecular processes behind the ARS in the bulk transcriptome\u003c/h2\u003e \u003cp\u003eTo explore the molecular pathways behind the correlation between ARS and NSCLC prognosis, we used functional enrichment analysis.\u003c/p\u003e \u003cp\u003eBased on the GO gene set, we performed GSEA analysis and found that the high-risk group was enriched for positive regulation of cell proliferation, differentiation and apoptosis processes as well as energy metabolism processes, whereas the low-risk group was enriched for positive regulation of oxidative stress, bile acid and salt metabolism processes as well as characteristic immunity \u003cb\u003e(Fig.\u0026nbsp;7A)\u003c/b\u003e. Furthermore, as shown in \u003cb\u003eFig.\u0026nbsp;7B and Fig.\u0026nbsp;7C\u003c/b\u003e, the low-risk group showed higher activity in pathways related to HALLMARK_BILE_ACID_METABOLISM, HALLMARK_HEME_METABOLISM and HALLMARK_FATTY_ACID_METABOLISM, while the high-risk group showed stronger activity in pathways associated with IL6_JAK_STAT3_SIGNALING, HALLMARK_G2M_CHECKPOINT, MYC_TARGETS_V2, HALLMARK_HYPOXIA and HALLMARK_E2F_TARGETS. Correlation studies between ARS and Signature Pathway scores provided further support for these results, indicating a substantial association between ARS and metabolic pathways and biological processes associated with cancer.\u003c/p\u003e \u003cp\u003eWe used KM curve analysis to determine whether Hallmark pathways were associated with NSCLC prognosis. Poor prognosis was associated with pathways positively associated with ARS, including IL6_JAK_STAT3_SIGNALING ,and HALLMARK_HYPOXIA \u003cb\u003e(Fig.\u0026nbsp;7E,F)\u003c/b\u003e. On the other hand, prognosis was positively correlated with pathways inversely associated with ARS, including HALLMARK_E2F_TARGETS \u003cb\u003e(Fig.\u0026nbsp;7D,G)\u003c/b\u003e. These findings suggest that the different prognostic outcomes seen in ARS risk subgroups may be due to either activation or inhibition of these mechanisms.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eLandscape of genomic variation and tumour heterogeneity across ARS categories\u003c/h2\u003e \u003cp\u003eIntra-tumor hetero(ITH) geneity is a well-known genetic feature of cancer that arises from the accumulation of mutations (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). It has been shown that ITH is linked to a higher risk of cancer and treatment resistance (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Higher MATH scores were linked to higher ITH in this study, which assessed ITH in patients with NSCLC using the mutant allele tumour heterogeneity (MATH) approach. The high-risk group of NSCLC patients had higher MATH scores, as shown in \u003cb\u003eFig.\u0026nbsp;8A\u003c/b\u003e. Patients in the high risk\u0026thinsp;+\u0026thinsp;high MATH group had a considerably poorer prognosis than patients in the low risk\u0026thinsp;+\u0026thinsp;low MATH group when we combined ITH and NSCLC (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This suggests that using these two measures together may improve prognosis for NSCLC patients. \u003cb\u003e(Fig.\u0026nbsp;8B)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eTo examine the differences in genomic mutations across the ARS subgroups, we have shown the mutation patterns between the high and low risk groups in \u003cb\u003eFig.\u0026nbsp;8C,D\u003c/b\u003e. Different aspects of the mutations were observed between the two groups. For example, DNA repair and apoptosis are mediated by the important tumour suppressor gene TP53(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). The high-risk group saw a mutation frequency of 56% as opposed to 45% for the low-risk group. When MUC16, another tumour-promoting factor, is mutated, it can lead to immunological escape, chemotherapy resistance and a poor prognosis. The mutation frequency was 35% in the high-risk group and 40% in the low-risk group(\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). The differences in prognosis seen in the risk sub-groups of the ARS may be explained by the different mutational landscapes between the risk classes.\u003c/p\u003e \u003cp\u003eWe explored the link between co-occurring and exclusive mutations in the top 20 mutated genes in both the high and low risk categories. According to the results, there was a higher incidence of co-occurring mutations in the high-risk group \u003cb\u003e(Fig.\u0026nbsp;8E, F)\u003c/b\u003e. CNV of the top 11 genes with the largest changes between ARS risk categories was also examined \u003cb\u003e(Fig.\u0026nbsp;8G)\u003c/b\u003e. According to our results, the main changes in FLG, KRAS and CSMD3 were due to CNV gain, whereas the main changes in LRP1B, MUC16 and TP53 were due to CNV loss.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eThe connection between the anti-cancer immune cycle, immunotherapy response and the ARS\u003c/h2\u003e \u003cp\u003eIn order to evaluate the immune infiltration status of NSCLC samples, we used the ESTIMATE technique to calculate the immunological score, stroma score, ESTIMATE score, and tumour purity score for NSCLC risk groups. Figure\u0026nbsp;9A-C shows that the immunity, ESTIMATE, and tumour purity scores were considerably higher in the high-risk group. We also found that 12 anoikis-related genes had strong correlations with tumour infiltrating immune cells, with monocytes and FCGRT and FBP1 having good correlations and macrophage M2 having positive correlations with CTSH and NPC2 \u003cb\u003e(Fig.\u0026nbsp;9F)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eFor the purpose of analyzing the variations in particular immune cell infiltration between the high-risk and low-risk groups, we used the CIBERSORT technique to quantify the number of invasive immune cells in each sample \u003cb\u003e(Fig.\u0026nbsp;9D)\u003c/b\u003e. In addition, we validated the result with the ssGSEA algorithm and showed that the high-risk group had higher levels of CD8 T cells and memory B cells. \u003cb\u003e(Fig.\u0026nbsp;9E)\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eUsing Spearman\u0026rsquo;s correlation analysis, we then searched for immune cell types that were significantly associated with ARS and found 12 cell types (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) \u003cb\u003e(Fig.\u0026nbsp;9G)\u003c/b\u003e. Using overlapping Venn diagrams, we were finally able to identify seven overlapping TME celltypes :T_cells_CD4_memory_resting,T_cells_CD4_memory_activad, Monocytes, Macrophages_M0, Macrophages_M2, Dendritic cells_resting and Mast cells_resting \u003cb\u003e(Fig.\u0026nbsp;9H)\u003c/b\u003e. These findings highlight the significance of these seven forms of immune cell infiltration for the development and course of NSCLC.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eThe relationship between ARS and immunotherapy response\u003c/h2\u003e \u003cp\u003eLarge immune cell infiltrations alone are not sufficient to characterise immune activation and exhaustion due to the complexity of immunological processes and the environment within the tumour. Nevertheless, a more comprehensive comprehension of the anti-cancer mechanism of immune cells may be acquired by evaluating the efficacy of each phase of the immune cycle in combating cancer. This will provide a better direction for immunotherapy. Figure\u0026nbsp;10A shows that the ARS risk subgroups differed significantly in phases 2, 4, 5 and 7 of the anti-cancer immune cycle. Phase 7 showed that the high-risk group was more active in killing cancer cells. In addition, \"Step 4 - Immune cell trafficking to the tumour\" was further developed to investigate the recruitment of different immune cells by the ARS risk subgroups. The results showed that helper T cells, regulatory T cells and immune cells including CD16\u003csup\u003e+\u003c/sup\u003e monocytes and CD8\u003csup\u003e+\u003c/sup\u003e memory cells were more abundant in the high-risk group \u003cb\u003e(Fig.\u0026nbsp;10B)\u003c/b\u003e. These results suggest that people at higher risk have more anticancer activity in the immune cell function cycle.\u003c/p\u003e \u003cp\u003ePrevious studies have shown that increased levels of immune checkpoint expression are linked to a more favorable response to ICIs (\u003cspan additionalcitationids=\"CR30\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e). We therefore looked at immune checkpoint expression levels across ARS risk categories. The majority of immune checkpoints, including TNFRSF25, TIGIT, CTLA-4, PDCD1 (PD1) and LAG3, were significantly overexpressed in the high-risk group, as shown in \u003cb\u003eFig.\u0026nbsp;10C\u003c/b\u003e. To confirm our findings, we examined the IPS scores derived from the TCIA database. In four categories: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) ips_ctla4_pos_pd1_pos (CTLA4 + /PD1- treatment), (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) ips_ctla4_pos_pd1_neg (CTLA4 + /PD1- treatment), (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) ips_ctla4_neg_pd1_pos (CTLA4-/PD1-treatment), and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) ips_ctla4_pos_pd1_pos (CTLA4-/PD1-treatment), and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) ips_ctla4_neg_pd1_neg (CTLA4/PD1 treatment), higher IPS scores were associated with a better response to ICI treatment, which included PD-1 inhibitor and CTLA4 inhibitor treatment. Our findings indicate that the immune checkpoint inhibitors CTLA4 + /PD1\u0026thinsp;+\u0026thinsp;and CTLA4 + /PD1 - treatment resulted in a substantially greater IPS in the high-risk groups. This suggests that patients in the high-risk group responded more favourably to both anti-CTLA4 and anti-CTLA4 + /PD1 therapy and that the PD-1 and anti-CTLA4 combination therapy was better compared to the low-risk group \u003cb\u003e(Fig.\u0026nbsp;10D-G).\u003c/b\u003e\u003c/p\u003e \u003cp\u003eWe included the atezolizumab-treated IMvigor210 cohort to further validate the prognostic power of ARS in predicting patient response to immunotherapy. We determined the cohort's risk score using the ARS model and then divided the patients into high-risk and low-risk groups. The high-risk group had a significant increase in tumour mutational burden (TMB), a well-established indicator of response to immunotherapy \u003cb\u003e(Fig.\u0026nbsp;10H)\u003c/b\u003e. The percentage of complete remission/partial remission (CR/PR) was significantly higher in the high-risk group according to the chi-squared test, while the low-risk group had a higher number of stable disease/progression cases (SD/PD) \u003cb\u003e(Fig.\u0026nbsp;10I,J)\u003c/b\u003e. In addition, those with CR/PR had significantly higher risk scores compared to those with SD/PD \u003cb\u003e(Fig.\u0026nbsp;10K)\u003c/b\u003e. In summary, these results confirm the ARS's capability to forecast the effectiveness of immunotherapy and indicate that patients classified as high-risk will have greater advantages from the medication.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eExamination of the relationship between medication sensitivity and the ARS\u003c/h2\u003e \u003cp\u003eFirst-line treatment for advanced NSCLC often consists of cytotoxic drugs and multi-targeted tyrosine kinase inhibitors (TKIs)(\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). In order to do this, we investigated the sensitivity of the ARS risk subgroup to several cytotoxic drugs (cisplatin and paclitaxel) and tyrosine kinase inhibitors (ositinib, gefitinib and gemcitabine). According to our research, the half IC50 of gefitinib was much lower in the low-risk group, and there was a positive correlation between the risk score and the IC50 of gefitini. On the other hand, ositinib, gemcitabine, cisplatin and paclitaxel had lower IC50s for the high-risk group, and there was a negative correlation between risk scores and IC50s for these drugs \u003cb\u003e(Fig.\u0026nbsp;11)\u003c/b\u003e. The findings indicate that those classified as high-risk shown a more favorable response to the therapeutic regimen consisting of ositinib, gemcitabine, cisplatin, and paclitaxel. Conversely, those classified as low-risk exhibited a higher degree of sensitivity to gefitinib.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn order to determine consistent and trustworthy predictive characteristics for NSCLC, we first developed an exclusive computational framework including 10 advanced machine learning algorithms and 101 potential combinations. The results of our analysis revealed ARS, which outperformed other research attempting to create a prognostic classifier for NSCLC programmed cell death-related prognosis in terms of predictive accuracy and clinical translational importance. Second, we risk stratified NSCLC patients using the ARS and evaluated their response to immunotherapy and sensitivity to first-line NSCLC drugs such as cytotoxic and TKIs. These results provide logical recommendations for the use of immunotherapy and chemotherapy in clinical practice, a significant step towards a more successful personalised medicine strategy. Furthermore, we have discovered potential medications that impede the advancement of NSCLC towards high-risk characteristics, presenting novel insights into preventative strategies for the illness. In addition, to gain a better understanding of ARS, we performed extensive multi-omics analyses, including bulk transcriptome and genome analyses, in contrast to other research that only examined the predictive power of the features. Our studies have identified the underlying processes and molecular underpinnings at different histological stages, providing evidence for the substantial correlation between ARS and NSCLC progression and prognosis. These findings also provide biological rationale and evidence for ARS that may guide personalised treatment strategies. Finally, we used a unique bioinformatics strategy integrating AUCell, UCell, GSVA, Singscore, AddModuleScore, ssGSEA and WGCNA algorithms to discover ARS at the bulk transcriptome level. Using this method, we were able to identify genes associated with ANO that could serve as potential targets for therapeutic intervention in NSCLC specifically. Furthermore, these discoveries offer fresh perspectives for future investigations of ANO in NSCLC.\u003c/p\u003e \u003cp\u003eWe have identified genes specifically associated with anoikis in NSCLC patients and used these genes to build a predictive model. One of the identified proteins, FBP1, has been identified as a tumour suppressor that is missing in many cancers. FBP1 functions as a protein phosphatase and plays an important role in inhibiting cancer progression(\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). According to a study by Emilie Dalloneau et al, global expression of FCGRT mRNA may indicate the abundance of antigen-presenting cells and the immune response to tumours in NSCLC. This finding could potentially help the decision-making process for patients with NSCLC(\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e). MT2A is responsible for the production of metallothionein 2A protein. Expression levels of MT2A mRNA are considered an indicator of poor prognosis in lung cancer patients. In addition, they have the ability to suppress MT2A expression, leading to cell death and apoptosis in tumour cells(\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). NUPR1 is a nuclear transcriptional regulator that can be induced to be expressed under various stressful environments and conditions and is involved in a variety of cellular physiological processes. Xiaotong Guo et al. report that silencing NUPR1 by tail vein injection of lentivirus-encoded shRNA also inhibited tumour growth in vivo. In addition, lentivirus-mediated RNAi targeting of NUPR1 significantly decreased the growth of NSCLC cells and triggered apoptosis in vitro (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e). According to the findings of Ji Young Kim et al, ARRB2 was found to be functionally involved in both signalling axes (TRAF6-TAB2 and TRAF6-BECN1 signalling pathways) in response to TLR3/4-stimulated NF-κB activation and autophagy induction to orchestrate lung cancer progression(\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e). In addition, targeting RGS1 may increase the efficacy of immunotherapy, as it has been shown to be the highest RGS family gene positively associated with immunogenicity. Baojun Wang et al. describe RGS1 as a potential target for immunotherapy(\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e). HLA-DQA1 is typically expressed on antigen-presenting cells and is part of the human leukocyte antigen (HLA) complex. It is required for leukaemia immunotherapy and has been shown to enhance immune responses(\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn this study, we emphasize the value of ARS in directing personalized therapy and targeted prevention in NSCLC, which may assist improve patient outcomes and save wasteful treatment expenditures. Overall, ARS may be a useful tool to provide physicians with the essential data for individualised drug selection. However, there are several drawbacks to this study. First, although we assessed and verified ARS in the training and validation datasets, further multicentre prospective investigations are needed to confirm our results. Secondly, further in vivo and in vitro research is needed to clarify the biological roles of ARS in NSCLC. Finally, although we hypothesised how sensitive different small molecule drugs would be to ARS risk categories, further research using in vitro drug testing and clinical trials is needed to confirm our findings.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this work, we developed an anoikis-related signature that might be a helpful tool for prognosis prediction, prevention and personalised therapy of NSCLC patients. In addition, we provided new insights from the fields of bulk transcriptomics and genomics into the molecular pathways underlying the initiation and development of NSCLC.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contribution Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eYuqi Ma\u003c/strong\u003e: Writing \u0026ndash; original draft, Project administration, Methodology, Formal analysis, Data curation, Conceptualization.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJia Li\u003c/strong\u003e: Writing \u0026ndash; original draft Conceptualization, Methodology，Visualization and formal analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTao Shen\u003c/strong\u003e: Review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe methods of analysis and the packages used are described in the \u0026quot;Materials and methods\u0026quot; section. All other codes are available from the corresponding author upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data achieved and analyzed in the current study are available in the TCGA repository (https://portal.gdc.cancer.gov/) and GEO database (https://www. ncbi.nlm.nih.gov/geo/).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Competing Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors voluntarily participated in this study and declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: a cancer journal for clinicians. 2021;71(3):209\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchabath MB, Cote ML. Cancer Progress and Priorities: Lung Cancer. Cancer epidemiology, biomarkers \u0026amp; prevention: a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 2019;28(10):1563\u0026ndash;79.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaron EB, Hellmann MD, Rizvi NA, Carcereny E, Leighl NB, Ahn MJ, et al. Five-Year Overall Survival for Patients With Advanced Non\u0026ndash;Small-Cell Lung Cancer Treated With Pembrolizumab: Results From the Phase I KEYNOTE-001 Study. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 2019;37(28):2518\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGout DY, Groen LS, van Egmond M. The present and future of immunocytokines for cancer treatment. Cellular and molecular life sciences: CMLS. 2022;79(10):509.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRibas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade. Science (New York, NY). 2018;359(6382):1350\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeeke S, Hofman P. Tumor mutational burden assessment as a predictive biomarker for immunotherapy in lung cancer patients: getting ready for prime-time or not? Translational lung cancer research. 2018;7(6):631\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang D, Zhou X, Zhang K, Yu Y, Cui SW, Nie S. Glucomannan from Aloe vera gel maintains intestinal barrier integrity via mitigating anoikis mediated by Nrf2-mitochondria axis. International journal of biological macromolecules. 2023;235:123803.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhong X, Rescorla FJ. Cell surface adhesion molecules and adhesion-initiated signaling: understanding of anoikis resistance mechanisms and therapeutic opportunities. Cellular signalling. 2012;24(2):393\u0026ndash;401.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYe G, Yang Q, Lei X, Zhu X, Li F, He J, et al. Nuclear MYH9-induced CTNNB1 transcription, targeted by staurosporin, promotes gastric cancer cell anoikis resistance and metastasis. Theranostics. 2020;10(17):7545\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBuchheit CL, Angarola BL, Steiner A, Weigel KJ, Schafer ZT. Anoikis evasion in inflammatory breast cancer cells is mediated by Bim-EL sequestration. Cell death and differentiation. 2015;22(8):1275\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChi H, Jiang P, Xu K, Zhao Y, Song B, Peng G, et al. A novel anoikis-related gene signature predicts prognosis in patients with head and neck squamous cell carcinoma and reveals immune infiltration. Frontiers in genetics. 2022;13:984273.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun Z, Zhao Y, Wei Y, Ding X, Tan C, Wang C. Identification and validation of an anoikis-associated gene signature to predict clinical character, stemness, IDH mutation, and immune filtration in glioblastoma. Frontiers in immunology. 2022;13:939523.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMariathasan S, Turley SJ, Nickles D, Castiglioni A, Yuen K, Wang Y, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature. 2018;554(7693):544\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai Z, Zhou F. A novel Anoikis and immune-related genes marked prognostic signature for colorectal cancer. Medicine. 2022;101(46):e31127.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiao X, Guo C, Li S. Identification of a novel anoikis-related gene signature to predict prognosis and tumor microenvironment in lung adenocarcinoma. Thoracic cancer. 2023;14(3):320\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026auml;nzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC bioinformatics. 2013;14:7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research. 2015;43(7):e47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa D, Jiang YZ, Liu XY, Liu YR, Shao ZM. Clinical and molecular relevance of mutant-allele tumor heterogeneity in breast cancer. Breast cancer research and treatment. 2017;162(1):39\u0026ndash;48.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMroz EA, Tward AD, Hammon RJ, Ren Y, Rocco JW. Intra-tumor genetic heterogeneity and mortality in head and neck cancer: analysis of data from the Cancer Genome Atlas. PLoS medicine. 2015;12(2):e1001786.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMroz EA, Rocco JW. MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma. Oral oncology. 2013;49(3):211\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRajput A, Bocklage T, Greenbaum A, Lee JH, Ness SA. Mutant-Allele Tumor Heterogeneity Scores Correlate With Risk of Metastases in Colon Cancer. Clinical colorectal cancer. 2017;16(3):e165-e70.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeeleher P, Cox N, Huang RS. pRRophetic: an R package for prediction of clinical chemotherapeutic response from tumor gene expression levels. PloS one. 2014;9(9):e107468.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang C, Zhang H, Chen M, Wang S, Qian R, Zhang L, et al. A survey of optimal strategy for signature-based drug repositioning and an application to liver cancer. eLife. 2022;11.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Jr., Kinzler KW. Cancer genome landscapes. Science (New York, NY). 2013;338(6127):1546\u0026ndash;58.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nature reviews Clinical oncology. 2018;15(2):81\u0026ndash;94.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang J, Zhao X, Wang Y, Liu T, Zhong C, Lao Y, et al. Genomic alterations driving precancerous to cancerous lesions in esophageal cancer development. Cancer cell. 2023;41(12):2038-50.e5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang XY, Hong LL, Ling ZQ. MUC16: clinical targets with great potential. Clinical and experimental medicine. 2024;24(1):101.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShah A, Chaudhary S, Lakshmanan I, Aithal A, Kisling SG, Sorrell C, et al. Chimeric antibody targeting unique epitope on onco-mucin16 reduces tumor burden in pancreatic and lung malignancies. NPJ precision oncology. 2023;7(1):74.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu-Monette ZY, Zhou J, Young KH. PD-1 expression and clinical PD-1 blockade in B-cell lymphomas. Blood. 2018;131(1):68\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQin S, Xu L, Yi M, Yu S, Wu K, Luo S. Novel immune checkpoint targets: moving beyond PD-1 and CTLA-4. Molecular cancer. 2019;18(1):155.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChauvin JM, Zarour HM. TIGIT in cancer immunotherapy. Journal for immunotherapy of cancer. 2020;8(2).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJ\u0026auml;nne PA, Baik C, Su WC, Johnson ML, Hayashi H, Nishio M, et al. Efficacy and Safety of Patritumab Deruxtecan (HER3-DXd) in EGFR Inhibitor-Resistant, EGFR-Mutated Non-Small Cell Lung Cancer. Cancer discovery. 2022;12(1):74\u0026ndash;89.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang L, Liu J. Immunological effect of tyrosine kinase inhibitors on the tumor immune environment in non-small cell lung cancer. Oncology letters. 2022;23(5):165.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu W, Chu H, Zhang Y, Luo T, Yu H, Zhu H, et al. Fructose-1,6-bisphosphatase 1 dephosphorylates IκBα and suppresses colorectal tumorigenesis. Cell research. 2023;33(3):245\u0026ndash;57.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Z, Li M, Jiang H, Luo S, Shao F, Xia Y, et al. Fructose-1,6-bisphosphatase 1 functions as a protein phosphatase to dephosphorylate histone H3 and suppresses PPARα-regulated gene transcription and tumour growth. Nature cell biology. 2022;24(11):1655\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang G, Tao J, Lin L, Qiu W, Lu Z. Repurposing FBP1: dephosphorylating IκBα to suppress NFκB. Cell research. 2023;33(6):419\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDalloneau E, Baroukh N, Mavridis K, Maillet A, Gueugnon F, Courty Y, et al. Downregulation of the neonatal Fc receptor expression in non-small cell lung cancer tissue is associated with a poor prognosis. Oncotarget. 2016;7(34):54415\u0026ndash;29.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWerynska B, Pula B, Muszczynska-Bernhard B, Gomulkiewicz A, Piotrowska A, Prus R, et al. Metallothionein 1F and 2A overexpression predicts poor outcome of non-small cell lung cancer patients. Experimental and molecular pathology. 2013;94(1):301\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYamasaki M, Nomura T, Sato F, Mimata H. Metallothionein is up-regulated under hypoxia and promotes the survival of human prostate cancer cells. Oncology reports. 2007;18(5):1145\u0026ndash;53.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim YJ, Hong SE, Jang SK, Park KS, Kim CH, Park IC, et al. Knockdown of NUPR1 Enhances the Sensitivity of Non-small-cell Lung Cancer Cells to Metformin by AKT Inhibition. Anticancer research. 2022;42(7):3475\u0026ndash;81.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim JY, Shin JH, Kim MJ, Kang Y, Lee JS, Son J, et al. β-arrestin 2 negatively regulates lung cancer progression by inhibiting the TRAF6 signaling axis for NF-κB activation and autophagy induced by TLR3 and TLR4. Cell death \u0026amp; disease. 2023;14(7):422.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang B, Jiang B, Du L, Chen W, Zhang Q, Chen W, et al. Tumor-intrinsic RGS1 potentiates checkpoint blockade response via ATF3-IFNGR1 axis. Oncoimmunology. 2023;12(1):2279800.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng L, Liu Q, Wang M, Gu Y, Wang J, Wei Q, et al. Genetic variants in the human leukocyte antigen region and survival of Chinese patients with non-small cell lung carcinoma. Carcinogenesis. 2020;41(9):1203\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerr W, Eichinger Y, Beshay J, Bloetz A, Vatter S, Mirbeth C, et al. HLA-DPB1 mismatch alleles represent powerful leukemia rejection antigens in CD4 T-cell immunotherapy after allogeneic stem-cell transplantation. Leukemia. 2017;31(2):434\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"NSCLC, single-cell, machine learning, anoikis","lastPublishedDoi":"10.21203/rs.3.rs-4640324/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4640324/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eNon-small cell lung cancer (NSCLC) is a prevalent form of lung cancer characterized by a significant death rate. Anoikis (ANO), refers to a distinct kind of programmed cell death that is strongly linked to the body's immune response to cancer. Nevertheless, the precise function of ANO in NSCLC is still not well understood.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eANO-related genes were analysed using multiple methods, including AUCell, UCell, single-sample gene set enrichment analysis (ssGSEA), Singscore, AddModuleScore, GSVA and weighted gene co-expression network analysis (WGCNA). We have developed an innovative machine learning framework that combines 10 different machine learning algorithms and 101 possible combinations of these algorithms. The goal of this framework is to build a reliable signature, known as the Anoikis-related signature (ARS), which is related to the phenomenon of anoikis. The performance of ARS was evaluated in both the training and validation sets. Column line graphs using ARS were developed as a quantitative technique to predict prognosis in clinical settings. Multi-omics studies, including genomic and bulk transcriptomic, were performed to gain more in-depth knowledge of prognostic features. We analysed the responsiveness of risk groups to immunotherapy and searched for tailored drugs to target specific risk categories.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eWe discovered 103 genes associated with ANO at both single cell and bulk transcriptome levels. A computational framework using machine learning and 101 combinations was used to generate the consensus ARS. This framework showed exceptional performance in accurately predicting prognosis and clinical change, and the ARS can also be used to predict the initiation, progression and spread of NSCLC. Statistical studies have shown that it is an independent prognostic determinant of (OS) and disease-specific survival (DSS) in NSCLC. The integrated column line graphs of the ARS provide an accurate and quantitative tool for clinical practice. We also identified distinct metabolic processes, patterns of genetic mutations and the presence of immune cells in the tumour microenvironment that differed between the high-risk and low-risk groups. Significantly, there were significant changes in the immunophenotype score (IPS) between the risk groups, suggesting that the high-risk group is likely to have a more favourable response to immunotherapy. In addition, potential drugs targeting specific at-risk populations were identified.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe purpose of our work is to create a signature associated with immunogenic cell death. This signature has the potential to be a useful tool for predicting the prognosis of NSCLC, as well as for targeted prevention and personalised therapy. We are also providing new insights into the molecular pathways involved in the growth and progression of NSCLC through the use of mass transcriptomics and genomics research.\u003c/p\u003e","manuscriptTitle":"Multi-omics analysis reveals an anoikis-related signature for non-small cell lung cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-23 10:52:35","doi":"10.21203/rs.3.rs-4640324/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9b429eb2-a648-4d3d-a835-f5891d0887a6","owner":[],"postedDate":"July 23rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":34779569,"name":"Biological sciences/Cancer"},{"id":34779570,"name":"Biological sciences/Computational biology and bioinformatics"}],"tags":[],"updatedAt":"2024-07-26T04:26:37+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-23 10:52:35","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4640324","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4640324","identity":"rs-4640324","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}